Feeding interactions between Atlantic salmon (Salmo salar) postsmolts and other planktivorous fish in the Northeast Atlantic

Fish samples from 161 pelagic trawl hauls were taken from nine different scientific surveys targeting salmon and other pelagic fish during spring and summer in 2008 and 2009. The surveys ranged from northwest of Ireland, to the central Norwegian Sea in June and July, and the northern Norwegian Sea in August (Fig. 1). The objective of these surveys was to sample postsmolts in the open ocean as part of the Salsea Merge project (http://salmonatsea.com/the-salsea-merge-project/), but mackerel and NSSH were also caught in many of the trawl hauls due to their pelagic nature. The trawl’s size selectivity of mackerel and NSSH is not known, but a range of individual sizes were caught, including large adult fish. The sampling did not follow standardized transects or equal distance spacing, but trawling was done over a large geographic area with more effort in areas with assumed high probability of postsmolt catch. More information about the surveys and biological sampling are presented in Table 1. Trawling was carried out at the surface at ∼4 knots (1 knot = 1.852 km·h^–1) with either a dedicated salmon trawl or another similar pelagic trawl. Two large floats attached to each side and a float in the center kept the trawl at the surface during towing. Once the fish were on deck, they were individually weighed to the nearest gram, and the total length was measured with a 5 mm resolution. The stomachs were removed from the fish and frozen. Salmon larger than 1000 g were excluded from the analyses to ensure the dataset only included postsmolts (the largest postsmolt being 622 g). The stomach contents of 750 postsmolts, 678 mackerel and 204 herring (Table 1) were later analyzed in the laboratory. The stomach content was identified to species level when possible or to closest family if it was too digested for species identification.

The number of individual prey items within each prey group was counted for all sampled stomachs. The weight of all unique prey groups in each stomach was recorded for Norwegian and Irish samples, but not for the Faroese samples. To correct for this missing information from the Faroese samples, the weight of each prey group was estimated based on the median individual prey item weight multiplied by the number of prey items. The median weight of all prey items was calculated based on all identified prey in Irish or Norwegian samples. When all prey in the stomach could be identified, the calculated weight of the prey group was summarized and scaled to the measured total weight of the stomach content. If there were unidentified prey in the stomachs, the difference between the total stomach content weight and sum of estimated prey group weights was assigned to this group. There were no prey species in Faroese samples that were not identified in the samples from the two other countries. Prey species and groups from each stomach were oven-dried separately at 70 °C for more than 24 h, and the dry weight was recorded.

Preliminary analyses revealed large spatial and temporal variation in the diet for all three species. The dataset was therefore split into four time periods due to the temporal and spatial coverage of the surveys. These periods were May, covering the areas northwest of Ireland, June covering the central Norwegian Sea, 1–15 July covering the southern and central Norwegian Sea, and 15 July – 10 August covering the central and northern Norwegian Sea. Data from samples collected in 2008 and 2009 were combined in the analyses.

For each period, the diet by weight was grouped into 12 different prey groups. Grouping was done based on preliminary analyses of the salmon diet. As a result of this, there are prey groups not relevant for NSSH or mackerel. Further, the group “Others” include prey organisms important for NSSH and mackerel, but not for postsmolts.

The diet overlaps between postsmolts and the two other species were calculated with Pianka’s index of niche overlap (Pianka 1974). This is a common index to evaluate whether two species have an overlapping diet (e.g., Langøy et al. 2012; Davis et al. 2015; Bachiller et al. 2018). The index is given bywhere the overlap index between the two species (O) is in the range 0–1. p_i,j and p_i,k are the proportions of the ith prey group by weight in the diets of species j and k, respectively. If O is 0 there is no diet overlap between the two species, while a value of 1 means a complete overlap. Partly overlapping diets is not necessarily equal to niche overlap, as species using resources independently of each other may still utilize some of the same resources. To test whether the species use the same resources more than what is expected by chance, we compared the overlap index to a null expectation (RA3 algorithm, 1000 repetitions) using the R package EcoSimR (Gotelli et al. 2015).

The feeding ratio (FR) was calculated to quantify potential difference in feeding intensity for the three fish species. This is an estimate of the weight of the stomach content relative to the fish weight and is a snapshot of the stomach content at the time of sampling. FR was calculated by the following equation:where m_f is the mass (g) of the fish and m_s is the mass (g) of the stomach content. A Mann–Whitney U test for nonparametric samples was used to test whether FR was significantly different between species.

Finally, as the three fish species can potentially feed on different sizes and (or) generations of prey, the most common prey organisms consumed by all three species were identified, and a Mann–Whitney U was applied to test whether the species preyed on the same size groups of prey.

Data on salmon postsmolts, herring (NSSH), and mackerel geographic distribution from three surveys (Table 1 — surveys 7, 8, and 9) were used to assess the geographic overlap among the species. For these surveys, fish density estimates with fine geographic resolution for postsmolts, mackerel, and NSSH in the central and northern Norwegian Sea had been calculated. For mackerel, density estimates were calculated from trawl catches, where catch per unit effort (CPUE) was calculated by dividing the total weight of mackerel with distance trawled. Acoustic estimates of NSSH density were used to assess overlap between postsmolts and NSSH. The acoustic data was sampled with a 38 kHz split-beam echo sounder. The survey area was gridded into a 2° longitude × 1° latitude grid, and the area’s respective average nautical area scattering coefficient (m²·nmi^–2; 1 nautical mile = 1.852 km) values were assigned to the trawl hauls within each grid cell.

Generalized additive models (GAMs; Wood 2006) were used with binomial error distribution to evaluate relationships between salmon presence–absence and spatial position, log-transformed herring acoustic backscatter, and log-transformed mackerel CPUE. Salmon presence–absence was used as the response variable instead of CPUE since the sampling design was not ideal for spatial modelling and because of the relatively low number of samples. Year was included as a random effect to account for differences between years that were not accounted for in the other variables, such as variations in sampling design. GAMs were fitted in the R library mgcv, using R version 3.5.3 (R Core Team 2019), and the approximate p values from the summary function and the shape of the smooth functions were used to evaluate spatial overlap among the species.

To explore whether interactions between postsmolts and pelagic fish can explain some of the temporal trends for salmon abundance, time series of survival and abundance of salmon were correlated to pelagic fish abundance. Time series of abundance of pelagic fish were obtained from two different sources, International Council for the Exploration of the Sea (ICES) assessment estimates of spawning stock biomass (SSB) (ICES 2018b, 2019b) and International Ecosystem Summer Survey in Nordic Seas (IESSNS) biomass index (Olafsdottir et al. 2018). There has been an expansion, and possibly a shift, in the geographic distribution of mackerel in recent years (Nøttestad et al. 2016). The survey index from the IESSNS trawl survey is a biomass estimate treated as an index due to unknown catchability of the survey and its gear. The survey is conducted annually in July–August and covers the Norwegian Sea, the area around Iceland, and southeastern parts of Greenland waters. The IESSNS survey index is included, as the ICES stock estimate may not fully represent the changing abundance of mackerel within the Nordic seas (Olafsdottir et al. 2018). The IESSNS survey index for the Norwegian Sea only, excluding the area around Iceland and Greenland, was also included, as the Norwegian Sea is a core feeding area for European postsmolts (Haugland et al. 2006). The data series for salmon included in the analyses are abundance and survival estimates for European salmon as provided by the ICES working group for North Atlantic Salmon (ICES 2018a) and estimated prefishery abundance (PFA) for Norway only (Anonymous 2018). Abundance estimates for Europe are PFA of 1 sea-winter fish (1SW) for the Northern and Southern Northeast Atlantic Commission (NNEAC and SNEAC; ICES 2018a). NNEAC is a geographic domain including the Scandinavian countries, Russia, and the northern part of Iceland. SNEAC includes UK, Ireland, France, Spain, and southern part of Iceland. Time series of estimated marine survival for five index rivers (ICES 2018a) distributed across Europe are also included. For more information about these time series see Table 2.

The time series for pelagic fish were correlated (Pearson’s correlation) with the time series for salmon. Time series were checked for stationarity and autocorrelation, as autocorrelated time series with strong trends increase the risk of spurious correlations (Shumway and Stoffer 2000). First, the nonstationarity was corrected by fitting a linear model with year as predictor to the original dataset. A detrended dataset was then obtained by subtracting the original dataset with predicted values from the linear model. Autocorrelation in the datasets were corrected by fitting AR(1) or AR(2) models to the datasets using R package Arima (Hyndman et al. 2019).

Stomachs from postsmolts and mackerel were sampled northwest of Ireland in May, but NSSH was not present in the area then (Fig. 2). The majority of mackerel sampled were juveniles weighing less than 100 g, which was similar to the weight of postsmolts sampled during this period (Table 1). Half of the diet by weight for postsmolts were fish larvae: 36% unidentified fish and 16% sand eel larvae. Other important prey were decapods (27%) and euphausiids (13%). The diet of mackerel consisted mainly of copepods and other zooplankton. Fish larvae made up nearly 10% of the mackerel diet, and euphausiids comprised around 6%. Pianka’s index for diet overlap between postsmolts and mackerel in this period and area was 0.22 and was not significant when compared with the null model (p = 0.54; Table 3).

Sampling in late June was conducted within a small geographic area in the central Norwegian Sea (66°N–67°N, 1°W–2°W). The postsmolts captured here had an average weight close to 100 g, while the mackerel and NSSH were on average more than 300 g (Table 1). All three fish species preyed upon copepods (Fig. 3). The proportion of Calanus finmarchicus by weight in the diet of mackerel and NSSH was 77% and 68%, respectively. This prey item was not a major component of the postsmolt diet, as it only made up 5% of the total stomach content. The dominating prey item in postsmolt stomachs was the copepod Anomalocera patersoni, which comprised 33% of the diet by weight. Further, Myctophidae larvae comprised 22% and undetermined fish 5% of the diet. These prey groups were not found in any of the mackerel or NSSH stomachs in this period. All three species also preyed upon amphipods and euphausiids, but these did not constitute a major part of the diet for any of the three species. These two prey groups comprised 25% of NSSH diet, 4% of the mackerel diet, and 11% of the postsmolt diet. Pianka’s index for the overlap in diet between postsmolts and mackerel in the same period was 0.22 (Table 3), which is identical to the overlap between postsmolts and NSSH in this period, as well as the diet overlap between postsmolts and mackerel in May. The diet overlap was not significant between postsmolts and mackerel (p = 0.31) nor between postsmolts and NSSH (p = 0.39).

Sampling in early July covered southern and central parts of the Norwegian Sea. The postsmolts had an average weight of more than 100 g, while mackerel and NSSH weighed on average more than 300 g (Table 1). The proportion of copepods in the diet had decreased for all three species compared with the samples from late June, but Calanus finmarchicus still made up 19%, 37%, and 1% of the diet for NSSH, mackerel, and postsmolts, respectively (Fig. 4). Amphipods were an important part of the diet for all three species, compromising 38% of the diet of postsmolts and 29%–30% for NSSH and mackerel. Euphausiids were also preyed upon by all three species, but they were more important for NSSH than the two other species. In addition, postsmolts preyed on Myctophidae, Sebastes spp., and unidentified fish larvae. These prey items were only found to a minor extent in mackerel or NSSH stomachs. The diet overlaps between postsmolts and the two other species in July was higher than what was observed in May and June. Pianka’s overlap index was 0.55 and 0.64 for postsmolts–mackerel and postsmolts–NSSH (Table 3), respectively. Although the diet overlap was higher than during the earlier periods, it was still not significant compared with the null models (p = 0.11 and p = 0.16 for mackerel and NSSH, respectively), which assume random feeding on available prey. Two postsmolts had pieces of plastic in their stomachs.

Sampling in late July and early August covered large parts of the Norwegian Sea. The postsmolt mean weight increased to between 100 and 150 g. The NSSH were adults with a mean weight of more than 300 g, but with smaller adolescent herring in the northeastern region (entrance of the Barents Sea). Mackerel were the same size as in early July, with a mean weight of more than 300 g. Amphipods were still the dominating prey in the postsmolt stomachs, comprising 54% of the stomach content by weight. NSSH stomachs contained 22% amphipods, but these only constituted 1% of the stomach content of mackerel (Fig. 5). Fish larvae remained an important part of the diet for postsmolts and was composed of 11% herring larvae, 15% unidentified fish larvae, and 6% Sebastes spp. larvae. Euphausiids were a dominant component of the diet for NSSH (32%), while the prey group “others” were dominating for mackerel (60%). Both of these groups were present in the postsmolt stomachs, but only as 6% and 7% of the diet by weight. Piankas diet overlap index was 0.56 between postsmolts and NSSH in late July and early August, while it was 0.14 between postsmolts and mackerel (Table 3). The diet overlap between postsmolts and NSSH was still not significant compared with the null model (p = 0.13) nor between postsmolts and mackerel (p = 0.52).

The lack of prey group weights for the Faroes samples made it necessary to estimate these weights based on the total stomach weight and prey group weights from Norwegian or Irish samples. The Faroese samples are approximately 1/4 of the total postsmolt dataset. The diet composition from the Faroese samples are similar to the Irish and Norwegian samples from the Norwegian Sea. However, for postsmolts the proportion of Sebastes and Myctophidae larvae is 5%–10% higher, while the proportion of herring larvae and Anomalocera patersoni is 5%–10% lower in the Faroese samples compared with the Irish and Norwegian data from the Norwegian Sea. For mackerel and NSSH, there are only small difference in the diet with the exception being that amphipods comprised nearly 30% of the diet in the Faroese mackerel stomachs compared with only 1% in the remaining mackerel stomachs sampled in the Norwegian Sea. Since the Faroese samples are taken in other parts of the Norwegian Sea, some deviation from the Norwegian and Irish samples is to be expected.

The diet during summer feeding in the Northeast Atlantic has previously been described for postsmolts, mackerel, NSSH, and blue whiting (e.g., Rikardsen et al. 2004; Prokopchuk and Sentyabov 2006; Haugland et al. 2006; Hvidsten et al. 2009; Langøy et al. 2012; Bachiller et al. 2016). However, there are few studies addressing the diet of postsmolts after the initial estuary and fjord entry. The novelty of this study lies in its analysis of diet of postsmolts and other pelagic fish sampled during the same surveys, where most of the samples come from trawl hauls catching at least two of the named species. The data presented from this study supports earlier findings for each of the species. Mackerel and NSSH mainly feed on various zooplankton species, with Calanus finmarchicus as an important part of the diet (Prokopchuk and Sentyabov 2006; Langøy et al. 2012; Bachiller et al. 2016). For postsmolts, the diet is mainly composed of fish larvae and amphipods (Rikardsen et al. 2004; Haugland et al. 2006). The relative abundance of the different prey organisms in the Northeast Atlantic is not known in detail, but C. finmarchicus constitutes a large part of the planktonic biomass in the Norwegian Sea during spring and summer (Melle et al. 2004). It is therefore likely that postsmolts selectively feed on fish larvae and a few zooplankton species, while mackerel and NSSH are more opportunistic in the prey search (Langøy et al. 2012). Selective prey search for postsmolts has previously been shown (Jacobsen and Hansen 2000; Salminen et al. 2001; Renkawitz and Sheehan 2011). Even in June in the Norwegian Sea, where C. finmarchicus most likely were abundant given the high proportion of this prey in NSSH and mackerel stomachs and the high abundance of C. finmarchicus generally observed in the Norwegian Sea at this time period (Bagøien et al. 2012), postsmolts barely consumed C. finmarchicus. Overall the diet overlap is low and never significant between postsmolts and the other two species. Many of the prey organisms are consumed by all three species, but in varying proportions. The exception being amphipods, which composed a large part of the diet for all three species in July. Although mackerel feed on smaller amphipods than postsmolts, they both feed on juvenile amphipods, and the minor difference in amphipod size may be due to small-scale geographic variation in size distribution or the filter feeding capability of mackerel (Macy et al. 1998). The results indicate that all three fish species mainly feed on juvenile Themisto (Noyon et al. 2011). The large proportion of amphipods in the diet in early July for all three species gives the highest diet overlap between postsmolts and mackerel or NSSH in this period. Euphausiids are another prey item consumed by all three species. However, it is not the dominant prey item for postsmolts in any of the sampled time periods. There are observations of unidentified fish larvae in the mackerel stomachs in July but not in June. We hypothesize that this is herring larvae, as herring is only identified from postsmolt stomachs sampled in July and the fact that mackerel have previously been reported to prey on herring larvae (Skaret et al. 2015). Mackerel in Icelandic waters feed on larvae of sandeel and mesopelagic fish (Óskarsson et al. 2016). Hence, mackerel and postsmolts prey on larvae of the same fish species. One sea-winter salmon and older salmon can prey on mackerel and herring in the Norwegian Sea (Jacobsen and Hansen 2001). The stomach samples presented here did not find evidence for mackerel predation on salmon postsmolts. Further, mackerel predation on salmon has not been mentioned in any previous publications presenting the diet of mackerel in the Northeast Atlantic (Langøy et al. 2012; Óskarsson et al. 2016; Skaret et al. 2015; Bachiller et al. 2016). The FR is not a measurement of whether the fish consume enough energy for growth and metabolism, but differences in FR between species can be an indication of interspecific competition. The FR is generally higher for postsmolts than for NSSH. However, NSSH start their feeding season before postsmolts and are more dependent on feeding in the spring (Varpe and Fiksen 2010). A lower FR during the summer for NSSH than for postsmolts is as expected, as salmon feed intensively throughout the summer (Haugland et al. 2006). Mackerel had a much higher feeding ratio than postsmolts in May, but this was south of the Norwegian Sea. Later in the summer, in the Norwegian Sea, postsmolts had a similar or higher FR as mackerel. Salmon have a faster somatic growth in the sea than mackerel. Thus, it is reasonable to assume that salmon need a higher daily energy intake than mackerel to achieve normal somatic growth. However, this can also be achieved by feeding on higher-energy prey or by lower energy utilization, for instance associated with metabolism.

Although the diet overlap is low, mackerel and (or) NSSH can potentially affect the prey availability for salmon due to their larger biomass. The biomass of prey consumed by the pelagic fish quickly exceeds the biomass consumed by salmon when there is an overlapping diet. The mackerel stomachs sampled in the Norwegian Sea consisted of 0%–2% fish larvae, while it was almost 10% in the mackerel stomachs sampled northwest of Ireland. These were lower than the proportion of fish larvae in postsmolt stomachs. However, there are many more mackerel compared with postsmolts feeding in the Northeast Atlantic. If ∼50% of the mackerel stock feed in the Norwegian Sea (Olafsdottir et al. 2018), there are now more than 2 million tonnes of mackerel in the postsmolt feeding areas. Further, assuming the postsmolts are 120 g, 3 500 000 adult fish return annually to their home rivers, a survival rate of 5%, and they all feed in the Norwegian Sea, there are <10 000 tonnes of postsmolts. The abundance of mackerel, NSSH, and blue whiting are orders of magnitude larger than that of postsmolts (ICES 2018b), and so is the total consumption. NSSH, mackerel, and blue whiting are estimated to consume 100–150 million tonnes of prey annually (Bachiller et al. 2018), but this also includes feeding outside of the Norwegian Sea ecosystem. Hence, even with a low diet overlap, interactions with pelagic fish may potentially affect the feeding conditions for postsmolts. However, according to Skjoldal et al. (2004), the pelagic fish are not assumed to be the most important predators on zooplankton in the Norwegian Sea ecosystem. Amphipods are estimated to consume nearly 200 million tonnes of prey annually, mainly smaller zooplankton (Skjoldal et al. 2004). Mesopelagic fish and the squid Gonatus are also important predators, with estimated annual consumption of 110 and 45 million tonnes wet weight prey, respectively (Skjoldal et al. 2004). Changes in biomass or spatial distribution of the key invertebrate predators in the Norwegian Sea ecosystem can most likely affect marine feeding conditions for both salmon and other pelagic fish.

The spatial overlap between postsmolts and NSSH is low. The probability of observing postsmolts did not increase with NSSH density. Hence, postsmolts and NSSH do not tend to aggregate in the same geographic areas. Further, the vertical overlap is most likely low since NSSH are normally located deeper than postsmolts (Huse et al. 2012; Holm et al. 2000). Postsmolts were found in the northern part of the Norwegian Sea, while adult NSSH were found farther southwest during the summer (ICES 2009). There was, however, a geographic overlap between postsmolts and juvenile NSSH in part of the distribution area of the postsmolts. Although the spatial modelling did not indicate that postsmolts and NSSH aggregated in the same geographic areas in 2008 and 2009, the geographic overlap between NSSH and postsmolts could have been greater in earlier periods, as the NSSH stock was found farther east and north in the Norwegian Sea in July–August during the 1990s and early 2000s (Utne et al. 2012). Mackerel, on the other hand, tend to aggregate in the same areas as postsmolts. The analyses presented here using data from 2008 and 2009 showed that mackerel are not found as far north as postsmolts, and there are large areas in the central and southern Norwegian Sea where mackerel are abundant but postsmolts were absent (Fig. 7b). However, a northward expansion of the mackerel stock after 2008–2009 (Nøttestad et al. 2016) has most likely resulted in a higher geographic overlap between postsmolts and mackerel in the past decade. Both species are located close to the surface during the summer (Nøttestad et al. 2016; Holm et al. 2000). However, the differences in diet during late June could indicate a small difference in feeding depth, as postsmolt feed on A. patersoni, a copepod living close to the surface, whereas mackerel fed mainly on C. finmarchicus, which are found in shallow depths but not directly at the surface.

Interactions with the NSSH stock have previously been suggested as a potential explanatory variable for changes in salmon abundance (Haugland et al. 2006), and the time series analysis presented here shows a negative correlation between NSSH abundance and PFA for NNEAC, which could potentially be due to feeding interactions. However, the result must be treated with caution, as data from 2000 and 2007 greatly affected the result, and the correlation is not significant when these years are excluded. These years are not classified as outliers and are therefore not excluded from the final calculations. The limited spatial and dietary overlap between postsmolts and NSSH reduce the probability of strong population regulating mechanisms between the two species. Postsmolts have a higher spatial overlap with mackerel, and interactions between these two species should therefore be expected to be higher than between postsmolts and NSSH. The time series for salmon PFA and survival were correlated against both stock estimates and survey indices of mackerel abundance, without revealing any significant negative correlation for any of the time series. The positive correlations between mackerel SSB and salmon PFA in Norway and between the IESSNS index and marine survival for salmon from the river Corrib in Ireland cannot be explained by feeding interactions. Perhaps bottom-up processes are affecting both postsmolts and mackerel, but this link is beyond the scope of this study. Overall, the lack of correlation between the time series of abundance of pelagic fish and salmon indicates that feeding interactions between Atlantic salmon postsmolts and other pelagic species in the Norwegian Sea are not the main drivers for the observed reduction in salmon PFA and marine survival.

There are several sources of uncertainty within the datasets used and also within the analyses. The digestion rate for fish larvae is faster than that for copepods (Hallfredsson et al. 2007). The proportion of fish larvae in the diet may therefore have been underestimated for all three fish species in this study. Sampling of fish was only done with surface hauls, and both diet and stomach fullness can change with vertical distribution of the fish. This is especially the case for NSSH, which also feed at several hundred metres depth (Nøttestad et al. 2007). Further, the lack of prey group weights for the Faroes samples made it necessary to estimate these weights based on the total stomach weight and prey group weights from Norwegian or Irish samples. The comparison of stomach samples from the Norwegian Sea from different nations showed that the lack of prey weights from Faroese samples did not have a large effect on the presented results. Testing the resource overlap against a null model using weights can have limitations, as the models do not test for independence or spatial and temporal sources of variation. Further, testing by prey weight can give a biased estimate of selectivity due to size differences between prey (Araùjo et al. 2011). We included these analyses in the study, as the size difference between fish larvae and zooplankton were relatively small, as stomachs of several species were sampled from the same trawl hauls, and trawl hauls were normally taken over a larger geographic area.

It is generally difficult to make firm conclusions concerning large-scale marine processes with high temporal and spatial variability. The mechanisms affecting spatial distribution and marine survival for postsmolts or for mackerel and NSSH are still not understood. However, potential feeding interactions resulting in reduced growth and survival of postsmolts could be revealed by indices such as diet overlap, spatiotemporal overlap, and negatively correlated abundance series. Postsmolts prey on larvae of herring and blue whiting (Haugland et al. 2006; Rikardsen and Dempson 2011), while 1SW and older salmon feed partly on both juvenile and adult pelagic fish (Jacobsen and Hansen 2001). In the Pacific Ocean, large jack mackerel (Trachurus symmetricus) have been observed to consume salmon smolts (Emmett and Krutzikowsky 2008). Mackerel predation on salmon postsmolts in the Northeast Atlantic can potentially happen as mackerel partly feed on small fish opportunistically (Iversen 2004). However, some species of Pacific salmon have smaller smolt than Atlantic salmon, making the smolt vulnerable to smaller predators such as jack mackerel. No salmon postsmolts were observed in the 678 mackerel stomachs presented here, nor have other publications of mackerel diet in the Northeast Atlantic reported such findings (Langøy et al. 2012; Óskarsson et al. 2016; Skaret et al. 2015; Bachiller et al. 2016). Postsmolts and mackerel or herring can compete directly for prey in areas or periods with limited prey or indirectly through trophic regulation ultimately affecting prey species. Fish larvae and carnivorous zooplankton often feed on C. finmarchicus, which is the most important prey for mackerel and NSSH. Pelagic fish may therefore potentially affect postsmolt feeding through competitive foraging. The species may also negatively affect each other due to aggregation in areas, causing a high fish density, in turn attracting predators that otherwise would migrate elsewhere. However, all the mentioned interactions should lead to a negative correlation between the abundance of pelagic fish and postsmolt if the interactions were strong. In our study, the diet overlap between postsmolts and NSSH or mackerel is low, and postsmolts have a low geographic overlap with NSSH. However, postsmolts overlap geographically with mackerel. Further, there is a weak to no correlation between the abundance or survival of salmon from key index rivers and the abundance of pelagic fish. In conclusion, it has not been shown in this study that the observed temporal changes in marine growth and survival of salmon is a result of feeding interactions with pelagic fish. It should, however, be emphasized that biological interactions occurring over large geographic areas and with large spatial and temporal variation can seldom be fully understood from biological sampling alone. Even though there seems to be limited interactions between postsmolts and pelagic fish in the Northeast Atlantic in general, there may still be local feeding interactions with pelagic fish negatively affecting postsmolts from certain regions in some years.